Recidivism is the most widely used outcome criterion for measuring treatment success in evaluation studies on offender treatment. Even though a reduction of offense severity would indicate a positive treatment effect, measures of severity are rarely used in research. In this paper, we present an instrument for measuring offense severity based on the statutory sentence lengths in German law. Our severity index was applied to a sample of N = 682 sex offenders and severity codings were carried out on criminal records in the German Federal Central Crime Register. Recidivism severity was defined as the highest severity score for an offense after release from prison. A correlation of r = .59 between the dichotomous recidivism criterion »severe recidivism« (any reconviction with a sentence length of two years or more, custody in a forensic hospital, or preventive detention) and recidivism severity scores support convergent validity. Bivariate analyses revealed significant correlations between offense severity scores and different variables that are associated with violent conduct. This supports the assumption that severity scores based on sentence length adequately reflect the relative severity of different offenses. Furthermore, we found significant correlations with the offenders’ age at the time of their release from prison and prior incarceration. This is in accordance with previous findings on correlates of offense severity. Our findings indicate that analyses of recidivism severity, for example the comparison of treated and untreated offenders, should be restricted to recidivists and complement analyses on binary outcome criteria. The application of our severity index was practicable and our results showed its utility in evaluation studies on offender treatment.